WorksPhysics and Astronomy
The anomalously large transiting planet around Boyajian's star: evidence for a circumplanetary ring
Abstract
A recent study reported a transiting giant-planet candidate with a mass of and a radius of on a wide orbit around Boyajian's star (KIC 8462852, also called Tabby's star), which is well known for its irregular, deep flux dips. The reported radius is puzzling, because cool giant planets of 10, are expected to have radii close to Jupiter's, so the companion appears 60–70% larger than standard models allow. Here we explore the solution that the planet is actually ringed. We develop a fast and accurate code to compute the transit light curve due to a ringed planet, and apply it to the TESS light curve. The ring model is marginally favoured by the data by and a difference in Bayesian information criterion (BIC) of . The preferred configuration is a Jupiter-sized () planet surrounded by an opaque, highly flattened ring extending to – planetary radii with a nearly polar obliquity. While the ring geometry is uncertain, simple estimates suggest that a debris annulus extending out to the planet's Roche limit is needed to explain the large transit depth. The ring thus removes the radius anomaly altogether and is qualitatively consistent with the dusty, comet-shepherding environment inferred from the star's historic dimmings.
Introduction
KIC 8462852, better known as Boyajian’s star or Tabby’s star, is an otherwise ordinary F3 V main-sequence star that has become one of the most enigmatic objects in the Kepler field. During the four years of Kepler monitoring it exhibited a series of abrupt, aperiodic flux dips, reaching depths of 15% and 22% in events near days 792 and 1519 of the mission, which were far deeper than any planetary transit and had no detectable periodicity ([4]). A series of dipping events have also been observed after the Kepler era (e.g., [5, 3]). The dips are chromatic, with stronger attenuation at shorter wavelengths, indicating occulting material of dust-sized grains rather than a solid body ([9]). A wide range of mechanisms has been proposed to explain the variability of Boyajian’s star, including disintegrating exocomets or comet swarms ([4, 14]), a family of Trojan bodies sharing the orbit of a massive perturber ([2]), debris from a recently consumed planet ([21]), a warped or clumpy circumstellar dust disk, and, after all, non-natural “megastructure” scenarios ([30]). None of these has gained universal acceptance, and the leading class of models still requires an unseen massive companion whose gravity shepherds the dust-producing bodies onto star-grazing orbits. A stellar companion to Boyajian’s star was discovered later ([23]), but with a separation au it is unlikely to be responsible for the mysterious light curve variation of Boyajian’s star.
Recently, C. Madurga-Favieres et al. [19] reported a single, isolated transit event in the Transiting Exoplanet Survey Satellite (TESS, [26]) photometry. The event is strikingly different from the irregular dips that have been seen before: it is symmetric to within the photometric precision, with a depth of and a duration of hours, exactly the morphology expected for a solid, roughly spherical body crossing the stellar disk. Combining the transit data with a re-analysis of archival radial velocities, C. Madurga-Favieres et al. [19] obtained a companion mass of and a radius of on an orbit with a period of at least days, with only five discrete period windows surviving the Kepler archival constraints. If confirmed, this companion would both constitute the first transiting giant planet/brown dwarf around Boyajian’s star. jian’s star and may be the long-sought perturber capable of driving the exocomet population responsible for the historic dimmings.
There is, however, a physical puzzle hidden in the numbers. Gas-giant planets more massive than Jupiter are supported largely by electron degeneracy pressure, and their radii are expected to be nearly mass-independent at cool temperatures. As shown in Figure 1, both planetary evolution models (e.g., J. J. Fortney et al. [11]) and the empirical mass-radius relation for cool giants (D. P. Thorngren et al. [28]) predict – across –, with an intrinsic dispersion of only . Inflation by stellar irradiation, which applies to close-in hot Jupiters, is negligible at the companion’s equilibrium temperature of K. At an age of Gyr (L. A. Pearce et al. [23]), the star is no longer young for the residual heat of formation to keep the planet substantially puffed up either. The reported radius of is therefore surprisingly large by –.
Figure 1. Mass–radius diagram of known transiting exoplanets with mass and radius measured to better than 50% uncertainty, taken from the NASA Exoplanet Archive (J. L. Christiansen et al. [6]). Planets with orbital periods d and d are shown in gray and green, respectively, and Solar System planets are also marked. The black curve with the gray band is the empirical mass–radius relation for cool ( K) giants (D. P. Thorngren et al. [28]), and the orange band shows the theoretical relation of giant planets of 1 Gyr at 1 au from J. J. Fortney et al. [11]. The recently reported planetary companion to Boyajian’s star by C. Madurga-Favieres et al. [19], denoted here Boyajian’s star b, is shown as red square. Its radius is – larger than what is expected for such a cool, massive planet.
In this paper we propose and quantify a geometric solution to the puzzle by adding a ring to the planetary companion. An opaque ring surrounding a much smaller planet blocks additional starlight, mimicking an anomalously large planetary radius. Similar ideas have been proposed to explain other planets. In particular, A. L. Piro & S. Vissapragada [24] showed that several of the lowest-density “super-puff” planets could in principle be explained by small rocky planets with opaque rings, and ringed configurations have also been invoked for individual ambiguous systems (e.g. M. Aizawa et al. [1]). In the case of Boyajian’s star, a qualitatively similar idea was advanced by F. J. Ballesteros et al. [2], who interpreted the historic, asymmetric 15% dip at D792 as the grazing transit of a Saturn-like ringed giant, with the later dimmings caused by a swarm of Trojan bodies sharing its orbit. The ring model in F. J. Ballesteros et al. [2] is only schematic, because no ring parameters were fitted, the light-curve reproduction was qualitative, and the predicted repeating events, with Trojan passages in 2021 and a new transit in 2023, have not been reported.
Our work differs from F. J. Ballesteros et al. [2] in several essential ways. First, the motivation of our ring model is to explain the abnormally large radius of the recently found transiting object, rather than to explain the overall light curve variations. As such, we aim to explain an actual, unambiguous transit event of the companion, whose orbital period and mass have been independently anchored by radial velocities (C. Madurga-Favieres et al. [19]). Unlike in F. J. Ballesteros et al. [2], we combine detailed light curve modeling with physical constraints to reveal the ring geometry. As shown later in the paper, our reported ring structure is also quite different from that proposed in F. J. Ballesteros et al. [2].
This paper is organized as follows. § 2 describes the TESS light curve and our standard transit fit; § 3 describes the new code for efficient computation of the transit light curve by ringed planet, building on the core algorithm of Q. Liu et al. [18]; in § 4, we present the ring model fit to the TESS light curve; finally, we briefly discuss our findings in § 5.
TESS data and standard transit fit
TESS light curve
We analyzed the TESS light curve of KIC 8462852 (Boyajian’s star) obtained during Sector 15 (2019 August 15 – September 10), which covers the single-transit event reported by C. Madurga-Favieres et al. [19]. We used the pre-search data conditioned (PDC) light curve at the short-cadence (2-min) sampling produced by the Science Processing Operations Center (SPOC) pipeline (J. M. Jenkins et al. [13]), retrieved from the Mikulski Archive for Space Telescopes (MAST). All times are given in Barycentric TESS Julian Date (BTJD = BJD − 2,457,000). All light-curve processing was performed with custom python routines built on lightkurve (Lightkurve Collaboration et al. [16]).
At the given transit event identified by C. Madurga-Favieres et al. [19] on 2019 September 3–4, we cut a d segment centered on the transit midpoint, masked the transit with a full window of 1.0 d, and fitted a quadratic polynomial to the out-of-transit points only. The entire segment, including transit event, was divided by this polynomial. The final product is a detrended, normalized light curve with the local baseline flat at unity and the full ingress/egress profile preserved, sampled at the native 2-min cadence, with time given in days relative to the adopted transit center (BTJD ). The standard deviation of the normalized out-of-transit light curve is taken as the flux uncertainties.
Standard transit fit
For consistency check, we have performed the standard K. Mandel & E. Agol [20] model fit to the flattened light curve. The model parameters are the mid-transit time , the impact parameter , the planet-to-star radius ratio , the quadratic limb-darkening coefficients (D. M. Kipping [15]), and the stellar density , which sets the orbital scale at fixed period. Because a single transit does not constrain the orbital period or the orbital shape, we fixed d, and . We used uniform priors over physically motivated ranges for all free parameters, with the exception of the limb-darkening coefficients. For and , we instead adopted Gaussian priors centered on the theoretical values (0.26 and 0.19, respectively) from A. Claret [7], evaluated at the spectroscopic stellar parameters of Boyajian’s star (T. S. Boyajian et al. [4]), with standard deviations of 0.10. We sampled the posterior with the affine-invariant MCMC ensemble sampler emcee (D. Foreman-Mackey et al. [10]), and the chains were run until the integrated autocorrelation times were much shorter than the discarded burn-in. The resulting parameters and associated uncertainties are listed in Table 1.
| Parameter | Spherical | Ring () | Ring () |
| Mid-transit time [d] | |||
| Impact parameter | |||
| Planet radius | 0.07 (fixed) | 0.07 (fixed) | |
| Ring inner radius | – | 0.07 (fixed) | 0.077 (fixed) |
| Ring outer radius | – | ||
| Ring flattening | – | ||
| Ring obliquity [deg] | – | ||
| [g cm] | |||
| (derived) | |||
| (derived) | |||
| Model ( d around ) | 644.7 | 631.2 | 630.4 |
| BIC | 0 |
Table 1. Best-fit parameters of the transit models for the TESS Sector 15 transit event of Boyajian’s star. Median values with the 16th and 84th percentiles are given for fitted parameters. The standard quadratic limb darkening parameters, and , are derived from the and parameters of D. M. Kipping [15]. The reference epoch of the detrended light curve is BTJD .
Our spherical-fit results are in good agreement with those of C. Madurga-Favieres et al. [19] despite the very different methodology. In particular, our radius ratio agrees within ( here vs. ), as does the impact parameter ( vs. ), and our inferred mean stellar density ( g cm) matches their joint-fit value ( g cm) within the errors. Consequently, the derived companion radius from our fit, , is consistent with their , confirming that it is unusually large. The reduced of the best-fit model is close to unity, although structured residuals near the transit bottom (Figure 2) hint at deviations from a perfect spherical object. As will be shown in § 4, they can be largely explained by the ring model.
Figure 2. Figure 2. Best-fit standard transit (i.e., spherical planet) and ring models to the TESS Sector 15 transit of Boyajian’s star. Top: sky-projected geometry of the best-fit ring model ( case), with the upper hemisphere of the limb-darkened stellar disk shown. The panel aligns vertically with the light-curve panels below (mid-transit at , ingress and egress at their respective epochs). The opaque system marks mid-transit, and the transparent systems mark ingress and egress (first and last contact of the model light curve). Middle: detrended 2-min TESS photometry (grey points; binned to 20 min with error bars) with the best-fit ring model (red; , , , ) and the best-fit spherical model (blue; , ) overplotted. Bottom: residuals of the spherical model and the difference between the ring and spherical model light curves (red), in parts per thousand (ppt). The ring redistributes occulted area to absorb the structured residuals near the transit bottom.
Implementation of the ring model in JoJo
The calculation of the transit light curve due to a ringed planet is non-trivial, especially for the purpose of MCMC modeling. Here we adopt the key algorithm of JoJo, first proposed in Q. Liu et al. [18] and used to model the transit light curve due to an oblate planet (Q. Liu et al. [17]), and apply it to the ring situation. 1 Our ring model includes a spherical body and a flat, circular ring viewed at an arbitrary obliquity. The ring is defined by its inner and outer radii ( and , in units of ), a flattening factor , the sky-projected obliquity of the ring’s major axis, and an effective opacity that controls the transparency of the ring. The aspect ratio of the projected elliptical shape of the ring is given by . A schematic plot of the ring model is illustrated in the upper right corner of Figure 3.
Figure 3. Comparison of the ringed-planet MCMC posteriors for two assumptions on the fixed inner ring radius: (blue) and (red). The dashed vertical lines mark in every panel involving . The two solutions are offset but overlapping: enlarging the inner radius moves the outer radius outward, increases the ring flatness and obliquity, and shifts the chord to smaller , illustrating the degeneracy between the ring geometry and the orbital configuration. The inset in the upper right corner illustrates the meaning of the model parameters related to the ring. Note that this cartoon geometry does not use the best-fit values of the ring model.
The occultation by the spherical planet follows the quadratic limb-darkened formalism of K. Mandel & E. Agol [20], while the ring contribution is obtained by integrating the stellar intensity over the overlap of the stellar disk with the elliptical annulus between the inner and outer ring edges. For each exposure the code solves the mutual intersections of the three bounding curves (stellar limb and the two ring ellipses, with the planet–ring intersections where the planet overlaps the ring hole) and evaluates the occulted flux through boundary integrals of the intensity along those arcs (Q. Liu et al. [18]), so all topological cases—ring fully on the star, ingress/egress, partial stellar limb crossings, and planet–ring overlap—are treated without heavy numerical grid integration. Limb darkening again uses the D. M. Kipping [15] triangular parameterization, and the orbit is parameterized as in Sect. 2.2.
We validate our ring code against PYPPLUS (E. Rein & A. Ofir [25]) and find it more efficient. To achieve the same precision, JoJo’s ring module is faster than PYPPLUS.
Ring model fit to the TESS light curve
Given the relatively high noise level of TESS data and the flexibility of the ring model, it is necessary to make a few reasonable assumptions about the ring model in order to make it physically plausible. Given the mass constraint on the transiting companion from C. Madurga-Favieres et al. [19], we can adopt the theoretically predicted radius according to the well understood mass–radius relation of cold giant planets (e.g., J. J. Fortney et al. [11]; see also Figure 1), , regardless of the assumed core mass. This gives .
The outer radius of the ring should be within the fluid Roche limit ([22])
where and are the mean densities of the planet and the ring particles, respectively. At the estimated separation of au, the planet is well within the water ice line of the F3V host, so the ring particles should be dominated by rocky rather than icy material. While both and are around the same order, their values may vary by a factor of a few. For example, the inferred mean density of the planet would be g cm with a mass of and a radius of , whereas the density of the ring particles can be as low as g cm (e.g., S. Watanabe et al. [29]; D. J. Scheeres et al. [27]). As so, we consider a ring outer radius highly unlikely, although a more quantitative upper bound is somewhat uncertain.
The inner ring radius is also uncertain. At equilibrium, the inner boundary is set by the balance between the supply due to collisional grinding and loss due to various drags, but its location cannot be determined from simple estimations in the lack of detailed physical properties of the planet (e.g., P. Goldreich & S. Tremaine [12]; J. N. Cuzzi et al. [8]). As will be shown later in the case of Boyajian’s star, of the ring model is not well constrained by itself, but a larger pushes out of the physically plausible range because the cross-sectional area should be conserved. For practical purposes, we therefore consider two quite extreme cases, and . This latter case matches Saturn’s ring system, with the D ring extending to times of the Saturn’s radius.
We assume a fully opaque ring with , and fix , , and in the same way as in the standard fit. For the limb darkening parameters, the same parameterization and priors as in the standard fit are adopted. We again use emcee to sample the posterior. The final run uses 64 walkers each with 10,000 steps, and the first 3000 steps are excluded from the posterior estimation.
Results from the ring model fits with two fixed are given in Table 1, and the posteriors are also illustrated in Figure 3. Note that the goodness-of-fit statistics ( and BIC) are evaluated on the 718 data points within d of the fitted mid-transit time, on the basis that the out-of-transit baseline has no contribution to the transit signal. Because the two ring models are nearly identical, we choose the one with and show its best-fit ring model along with the transit data and the best-fit spherical model in Figure 2. Compared to the standard spherical model, the ring solutions replace most of the planet’s cross-section with an extended, highly flattened annulus. With the planetary radius fixed at , the ring outer radius is –, corresponding to –, and the ring has a flatness of – and a projected obliquity of –. Assuming a concentric and circle ring, we can infer its inclination with respect to the line of sight –. The above values have relatively large uncertainties, because the ring-related parameters are not well constrained by the TESS data and there is a coherent degeneracy between parameters describing the ring geometry and those describing the transit geometry (e.g., the impact parameter and the orbital scale ), as shown in Figure 3. Nevertheless, our ring model fits suggest that the transit event of Boyajian’s star can be explained by a ring that is highly inclined with respect to the line of sight, nearly perpendicular to the transit chord, and spreading almost all the way to its Roche limit.
In terms of goodness of fit, both ring models improve the fits by relative to the spherical model, at the expense of two additional free parameters. The improvement is concentrated near the transit bottom, where the spherical model leaves structured residuals
(Figure Figure 2). After penalizing the extra parameters, the statistical evidence for the ring model is reduced, but it is still slightly favoured over the spherical model by (for ) and (for ). The ring model tries to explain the light curve feature at d offset from the mid-transit. Such a feature may be caused by other astrophysical effects such as spot crossing, which may not be rare for a relatively young and fast-rotating F3-type dwarf [4]. If so, what does it mean to the ring model? Excluding that light curve feature in the fits undoubtedly overturns the model comparison metric. However, because the standard transit model yields an unreasonably large size for the planetary-mass companion, the physical motivation for the introduction of the ring remains strong. From simple arguments, one can conclude that the ring, if exists, should be quite extended. Given the well constrained transit depth, the outer radius of the ring is , with the lower limit achieved if the ring is exactly face-on and connects all the way to the planet. In the more likely scenario of an inclined ring, , so a Saturn-like obliquity () would push the outer radius to , which is already close to the Roche limit (Equation (1)) and well consistent with the inferred values from our ring model.
Although an extended ring is required regardless of the detailed light curve morphology, its orientation is largely set by the feature at d after mid-transit. With , a highly oblique ring begins egress earlier than a spherical planet with the same cross-sectional area. Light curves of the two models begin to differ, and the difference between them therefore leaves a rebrightening feature, explaining the feature at d in the data.
With a nearly polar obliquity, the derived ring geometry is nevertheless intriguing. The true obliquity of the ring can be derived from its inclination and the projected obliquity , via , and the signs concern the ambiguity of the ring orientation. Adopting for both and , the true obliquity is or . Taking it as the spin direction of the planet, this super Jupiter is then rotating sideways, very similar to Uranus of our Solar System.
Discussion
We have presented a quantitative comparison of two interpretations of the single TESS Sector 15 transit of Boyajian’s star reported by C. Madurga-Favieres et al. [19]: a bare spherical planet, and a Jupiter-sized planet surrounded by an opaque ring. Both models reproduce the light curve at formally acceptable levels, and the ring improves the fit by for two additional parameters, concentrated at the structured residual near the transit bottom (Figure 2). After the penalty due to two more free parameters is taken into account, the two interpretations are statistically nearly indistinguishable, with to in favour of the ring model.
Although the statistical evidence for the ring is fairly marginal, the physical motivation for it is strong. The spherical fit requires a planet of at , which is – larger than the radii predicted by theoretical models and the empirical mass–radius relation (Figure 1). The ring fit absorbs most of the occulted area into a highly flattened, nearly polar annulus extending to – planetary radii, and thus the planetary radius is no longer abnormal.
If the ring is real, it has several implications for our understanding of the system as well as the ring formation in general. First, the ring extends nearly to the companion’s Roche limit. Given the location of the companion, it is a debris ring likely sustained by the disruption of comets or small satellites, which may be the same minor-body reservoir invoked to explain the historic dipping activity of this star. The ring component may also explain the brightening seen in the TESS data within d before and after the transit, which C. Madurga-Favieres et al. [19] also attribute to forward-scattering by possible circumplanetary dust. A detailed modeling of this feature in the context of our ring model is beyond the scope of the present work. If ringed, the companion would then simultaneously be the perturber that shepherds exocomets onto star-grazing orbits and a dusty object in its own right, making the system a scaled-up analogue of a ringed giant plus cometary swarm.
The nearly polar ring obliquity ( or ) implies that the ring plane is strongly tilted with respect to the orbital plane of the companion, a configuration reminiscent of Uranus and plausibly produced by the tidal disruption of a captured satellite or by obliquity excitation during the system’s dynamical history. We caution, however, that the ring orientation is only weakly constrained by the single available transit of relatively poor data quality, and that the apparent feature that drives the ring fit could in part be spot crossing (Sect. 4).
The ring and no ring interpretations make sharply different predictions for observations that directly probe the companion. The very direct test is precise photometry of the repeat transits, with the next expected in the 2026 October and 2027 January/March windows [19]. The ring and spherical
ends_mid=1 models differ most at ingress and egress, where the ring extends the duration of the partial phase and distorts its shape, and the predicted difference is at the few level, which is within reach of space-based time-series photometry of existing facilities, although the currently uncertain transit ephemerides makes the scheduling difficult. Depending on the composition and opacity of the ring particles, simultaneous multi-band photometry may also be able to test the ring model against the spherical planet model. Finally, with the upcoming Gaia DR4 epoch astrometry, the orbital properties and mass of the companion can be independently constrained (C. Madurga-Favieres et al. [19]). These measurements may provide conclusive evidence for the radius anomaly, even though they may not be able to test directly the ring model.
Acknowledgments
Work by W.Z. and Q.L. was funded by the National Natural Science Foundation of China (NSFC) under No. 12133005. This paper includes data collected by the TESS mission, which are publicly available from the Mikulski Archive for Space Telescopes (MAST). Funding for the TESS mission is provided by NASA’s Science Mission directorate.
References
- [1]Aizawa, M., Uehara, S., Masuda, K., Kawahara, H., & Suto, Y. 2017, AJ, 153, 193, doi: 10.3847/1538-3881/aa6336DOI
- [2]Ballesteros, F. J., Arnalte-Mur, P., Fernandez-Soto, A., & Martínez, V. J. 2018, MNRAS, 473, L21, doi: 10.1093/mnrasl/slx105DOI
- [3]Bourne, R., Gary, B. L., & Plakhov, A. 2018, MNRAS, 475, 5378, doi: 10.1093/mnras/sty097DOI
- [4]Boyajian, T. S., LaCourse, D. M., Rappaport, S. A., et al. 2016, MNRAS, 457, 3988, doi: 10.1093/mnras/stw218DOI
- [5]Boyajian, T. S., Alonso, R., Ammerman, A., et al. 2018, ApJL, 853, L8, doi: 10.3847/2041-8213/aaa405DOI
- [6]Christiansen, J. L., McElroy, D. L., Harbut, M., et al. 2025, PSJ, 6, 186, doi: 10.3847/PSJ/ade3c2DOI
- [7]Claret, A. 2017, A&A, 600, A30, doi: 10.1051/0004-6361/201629705DOI
- [8]Cuzzi, J. N., Burns, J. A., Chernazov, S., et al. 2010, Science, 327, 1470, doi: 10.1126/science.1179118DOI
- [9]Deeg, H. J., Alonso, R., Nespral, D., & Boyajian, T. S. 2018, A&A, 610, L12, doi: 10.1051/0004-6361/201732453DOI
- [10]Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306, doi: 10.1086/670067
- [11]Fortney, J. J., Marley, M. S., & Barnes, J. W. 2007, ApJ, 659, 1661, doi: 10.1086/512120DOI
- [12]Goldreich, P., & Tremaine, S. 1982, ARA&A, 20, 249, doi: 10.1146/annurev.aa.20.090182.001341DOI
- [13]Jenkins, J. M., Twicken, J. D., McCauliff, S., et al. 2016, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 9913, Software and Cyberinfrastructure for Astronomy IV, ed. G. Chiozzi & J. C. Guzman, 99133E, doi: 10.1117/12.2233418
- [14]Kiefer, F., Lecavelier des Etangs, A., Vidal-Madjar, A., et al. 2017, A&A, 608, A132, doi: 10.1051/0004-6361/201731306DOI
- [15]Kipping, D. M. 2013, MNRAS, 435, 2152, doi: 10.1093/mnras/stt1435DOI
- [16]Lightkurve Collaboration, Cardoso, J. V. d. M., Hedges, C., et al. 2018,, Astrophysics Source Code Library, record ascl:1812.013 http://ascl.net/1812.013
- [17]Liu, Q., Zhu, W., Masuda, K., et al. 2024, ApJL, 976, L14, doi: 10.3847/2041-8213/ad8f39DOI
- [18]Liu, Q., Zhu, W., Zhou, Y., et al. 2025, AJ, 169, 79, doi: 10.3847/1538-3881/ad98bc
- [19]Madurga-Favieres, C., Strøm, P. A., Wilson, T. G., Heidari, N., & Kiefer, F. 2026, MNRAS, 550, stag1273, doi: 10.1093/mnras/stag1273DOI
- [20]Mandel, K., & Agol, E. 2002, ApJL, 580, L171, doi: 10.1086/345520DOI
- [21]Metzger, B. D., Shen, K. J., & Stone, N. 2017, MNRAS, 468, 4399, doi: 10.1093/mnras/stx823DOI
- [22]Murray, C. D., & Dermott, S. F. 1999, Solar System Dynamics, doi: 10.1017/CBO9781139174817DOI
- [23]Pearce, L. A., Kraus, A. L., Dupuy, T. J., Mann, A. W., & Huber, D. 2021, ApJ, 909, 216, doi: 10.3847/1538-4357/abd33
- [24]Piro, A. L., & Vissapragada, S. 2020, AJ, 159, 131, doi: 10.3847/1538-3881/ab7192DOI
- [25]Rein, E., & Ofir, A. 2019, MNRAS, 490, 1111, doi: 10.1093/mnras/stz2556DOI
- [26]Ricker, G. R., Winn, J. N., Vanderspek, R., et al. 2014, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 9143, Space Telescopes and Instrumentation 2014: Optical, Infrared, and Millimeter Wave, ed. J. M. Oschmann, Jr., M. Clampin, G. G. Fazio, & H. A. MacEwen, 914320, doi: 10.1117/12.2063489
- [27]Scheeres, D. J., McMahon, J. W., French, A. S., et al. 2019, Nature Astronomy, 3, 352, doi: 10.1038/s41550-019-0721-3DOI
- [28]Thorngren, D. P., Marley, M. S., & Fortney, J. J. 2019, Research Notes of the American Astronomical Society, 3, 128, doi: 10.3847/2515-5172/ab4353DOI
- [29]Watanabe, S., Hirabayashi, M., Hirata, N., et al. 2019, Science, 364, 268, doi: 10.1126/science.aav8032DOI
- [30]Wright, J. T., Cartier, K. M. S., Zhao, M., Jontof-Hutter, D., & Ford, E. B. 2016, ApJ, 816, 17, doi: 10.3847/0004-637X/816/1/17